Research On The Fault Association Analysis For High-speed EMU

China's high-speed railway is developing rapidly.As a fast means of transport,high-speed EMU has been widely used,and its safe operation and maintenance work have become more and more important.With advanced data acquisition equipment and computer storage technology,China has generated and accumulated a large number of high-speed EMU fault information data as time goes by.How to extract effective information and knowledge from the mass data to instruct the repair work of the EMU and ensure the safe operation of the EMU is an urgent issue to be solved.Data mining can extract information of great importance hidden in the actual application data,that is,data mining can convert data into useful knowledge.As the main method of data mining,association rule mining can find the correlation between items in data transaction How to use the efficient association rule mining algorithm to mine the fault association from the fault information data of EMU to assist in the formulation of maintenance strategies and improve the safety and reliability of EMU is the focus of this thesis.The main work and innovations of this thesis include the following parts:(1)The current research status of EMUs and data mining in China and abroad are analyzed.The EMU repair theory and data mining related knowledge theory are summarized,and three association rule algorithms of Apriori,FP-Growth and Eclat are studied.(2)The EMU fault information data are sorted out,and the problems of EMU data are analyzed,then data preprocessing schemes such as data cleaning and data conversion are given;(3)This paper proposes an algorithm based on weighted thought of FP-Growth association mining called WFPAM algorithm by setting the weight to represent the importance of different items in the EMU fault data.In order to improve the efficiency and performance of the algorithm,the paper uses depth-first search method to traverse FP-tree in pre-order in order to avoid the problem of backtracking for several times when generating the conditional pattern base,and elaborates the idea of the improved algorithm.Then,the algorithm is parallelized and improved based on MapReduce programming model to be suitable for the mass data mining of EMU.Finally,the MR-WFPAM algorithm is put forward and the implementation process of the algorithm is designed;(4)FP-Growth,WFPAM and MR-WFPAM algorithms are used to mine the association of the EMU fault data.The performance of the three algorithms is compared and the result of experiments verifies the effectiveness and superiority of the improved algorithm.Finally,the excavated EMU fault association rules are analyzed and visualized.